Vision Research Center


News from the VRC

VRC Administration

VRC Facilities

Related Penn Sites


Vision Science Seminar


The Ideal Observer - ioprogs

Developed by Robert Smith - download

The ideal observer is a type of analysis that measures the
precision of a signal. To measure the precision of a neural
signal, a recording is made of the neuron's response to a pair
of repeating stimuli. Half of the responses are used to construct a probability histogram of the response amplitude, which defines the variability of the neural response to the stimuli. The other half of the responses are used to probe the histogram. The probability that a response was generated by one stimulus is compared with the probability that it was generated by the other stimulus. The one with the highest probability is chosen as the most likely stimulus, and the fraction of correct responses is calculated. When performed
for several pairs of stimuli varying e.g. in their contrast or
another parameter, this process generates a "neurometric function" that defines the neuron's ability to transmit information.

The "ioprogs" distribution contains several programs useful in
ideal observer analysis.

idisctmp template method, useful for responses with or w/o correlations.
idisc "joint probability" method, only useful for spike trains.
binread reads Axoscope format, converts to ASCII
crosscorr computes correlation between stimulus and response
genpot removes spikes to leave generator potential
spikerate computes instantaneous spike rate
poisson simple Poisson model of spiking
thresh finds threshold for spiking
voff sets voltage at beginning of trial to zero